🔬 optics

A re-entrant chip-free-space photonic interface for telecom-to-Rubidium spectroscopy

This paper presents a re-entrant photonic interface using a thin-film lithium niobate chip that converts telecom light to 780 nm, interacts with a Rubidium vapor cell in free space, and recollects the signal to achieve stable laser locking, thereby overcoming the limitations of traditional evanescent-field coupling for integrating external media with photonic integrated circuits.

Original authors: Jia-Lin Chen, Ruixin Zhou, Deng-Hong Liu, You-Long Fan, Zhu-Bo Wang, Min Chen, Xiang Fang, Jia-Qi Wang, Zheng-Fu Han, Guang-Can Guo, Ai-Ping Liu, Pengfei Wang, Xiaochi Liu, Juanjuan Lu, Wei Chen, Chan

Published 2026-07-17

📖 4 min read☕ Coffee break read

CC BY 4.0

Original authors: Jia-Lin Chen, Ruixin Zhou, Deng-Hong Liu, You-Long Fan, Zhu-Bo Wang, Min Chen, Xiang Fang, Jia-Qi Wang, Zheng-Fu Han, Guang-Can Guo, Ai-Ping Liu, Pengfei Wang, Xiaochi Liu, Juanjuan Lu, Wei Chen, Chang-Ling Zou

Original paper licensed under CC BY 4.0 (http://creativecommons.org/licenses/by/4.0/). ✨ This is an AI-generated explanation of the paper below. It is not written or endorsed by the authors. For technical accuracy, refer to the original paper. Read full disclaimer

Imagine a world where light is the ultimate delivery driver, zipping through tiny, super-efficient highways built right onto a computer chip. These "photonic integrated circuits" are like the microchips of the future, packing massive power into a space smaller than a fingernail. They are amazing at moving information, but they hit a wall when they need to talk to the messy, real world. Think of it like trying to have a deep conversation with a cloud of gas, a drop of blood, or a glowing liquid while keeping your head stuck inside a sealed, narrow tube. The light is trapped so tightly in the chip's tiny tunnels that it can't really reach out to touch these materials.

To solve this, scientists have tried a few tricks. Some try to squeeze the gas or liquid right onto the chip's surface, but the light only grazes the very edge, missing most of the material. Others send the light out of the chip, let it bounce around in a big, messy room full of mirrors and lenses, and then hope it comes back. But that big room is bulky, unstable, and defeats the whole point of having a tiny, portable chip. The big question in science right now is: How do we let our super-tiny chip light out to play with the big, free world, and then bring it back home without losing its way or needing a giant, clumsy setup?

This paper introduces a clever new solution called a "re-entrant chip-free-space interface." Think of it as a magical "emit-interact-recollect" loop. Instead of sending light out to a distant, separate machine, the chip itself acts like a smart lighthouse. It shoots a beam of light straight up into a small, self-contained module sitting just above it. This module holds the "guest" material—like a cloud of Rubidium atoms. The light zips through the cloud, does its job, hits a mirror on top of the module, and bounces straight back down. The chip then catches that returning light with a second, dedicated receiver, bringing the information right back into its tiny circuits for processing.

The researchers built a prototype using a special chip made of thin-film lithium niobate. Their goal was to turn invisible "telecom" light (the kind used for internet data) into visible red light that Rubidium atoms love. They successfully fired this red light up through a Rubidium vapor cell, let it bounce off a mirror, and caught it again on the same chip. The result? They were able to measure the atoms' "saturated absorption spectrum"—a fancy way of saying they could read the atoms' unique fingerprint with incredible precision. By locking their laser to this fingerprint, they kept the laser's frequency stable within ±280 kHz over a period of two hours.

What makes this so special is that they did it without needing a giant lab full of external mirrors and lenses. The entire process of generating the light, sending it out, interacting with the atoms, and catching it again happened in a loop that fits on a millimeter scale. The paper shows that this "re-entrant" design avoids the problems of previous methods: it doesn't squeeze the light too tight (which ruins the interaction), and it doesn't require the light to leave the chip entirely and get lost in the open air. Instead, it creates a perfect, closed loop where the external material becomes just another functional part of the chip itself. This opens the door to putting all sorts of things—like biological samples, other types of gases, or even laser gain media—onto a single chip in a way that was previously impossible, turning the chip into a truly universal platform for light and matter.

Technical Summary: A Re-entrant Chip-Free-Space Photonic Interface for Telecom-to-Rubidium Spectroscopy

Problem Statement
Photonic integrated circuits (PICs) offer high efficiency, scalability, and functional density by confining light in micron-scale waveguides. However, this tight confinement creates a fundamental limitation: on-chip modes cannot easily access or effectively interact with external media such as atomic vapors, gain media, or biological samples. Existing solutions face significant trade-offs. Directing light off-chip via fibers to separate free-space apparatuses sacrifices the compactness and stability of the PIC. Conversely, bringing media onto the chip to interact with weak evanescent fields restricts the interaction volume and induces severe transit-time broadening, degrading spectral resolution. Furthermore, approaches that emit a beam to interact with a medium and collect the result on an external detector fail to return the interacted light to the circuit for further on-chip processing.

Methodology
The authors propose and demonstrate a "re-entrant chip-free-space" interface that closes the optical loop by embedding external media directly into the PIC processing chain. The core architecture utilizes a "emit-interact-recollect" loop:

Emission: On-chip light is emitted into free space via a grating coupler (GC).
Interaction: The beam interacts with an external medium held in a thin module above the chip.
Re-collection: A reflector redirects the beam back through the medium, and a second, spatially separated GC on the same chip recollects the light.

To validate this concept, the team co-integrated a thin-film lithium niobate (TFLN) frequency-conversion chip with a Rubidium (Rb) vapor cell. The system operates as follows:

Frequency Conversion: Telecom-band light (1560 nm) is coupled into a periodically poled lithium niobate (PPLN) waveguide on the chip, where it undergoes second harmonic generation (SHG) to produce 780 nm light.
Free-Space Propagation: The 780 nm light is emitted vertically through a GC, passes through a Rb vapor cell, and is retro-reflected by a beam splitter.
Spectroscopy: The retro-reflected beam acts as a probe, passing through the cell a second time to enable saturated absorption spectroscopy (SAS).
Re-collection: The probe is recollected by a second GC (GC3) on the chip, routed through a passive waveguide, and detected by a photodetector.

The system employs two distinct GCs for emission and recollection to suppress direct back-reflection and stray-light interference, a limitation of shared-aperture designs. The geometry was optimized to ensure spatial overlap of the pump and probe beams within the cell while maintaining a millimeter-scale free-space span to minimize transit-time broadening.

Key Results

System Characterization: The TFLN chip successfully converted 1560 nm light to 780 nm with an on-chip power of ~230 mW, yielding ~0.9 mW of emitted 780 nm light. The beam profile was measured to be approximately 2.8 mm by 1.6 mm at a distance of 30 mm.
Spectral Resolution: The system resolved the hyperfine features of the $^{87}$ Rb D2 line, including crossover resonances. Despite a low total recollection efficiency (~0.32%) due to beam distortion and scattering, the signal-to-noise ratio was sufficient for spectroscopy.
Frequency Stabilization: Using the crossover resonance between the $F=2 \to F'=2$ and $F=2 \to F'=3$ transitions as a reference, the team implemented Pound-Drever-Hall (PDH) locking. The telecom laser was stabilized to the atomic resonance with a frequency drift of within ±280 kHz over 2 hours. This represents a significant improvement over the free-running drift of 190 MHz observed over the same period.
Transit-Time Broadening: Theoretical analysis and experimental data confirmed that the millimeter-scale free-space interaction keeps transit-time broadening at the ~1 MHz level, which is an order of magnitude below the intrinsic 87Rb linewidth (6 MHz), preserving spectral resolution.

Significance and Claims
The paper claims that this re-entrant interface establishes a general and scalable route to heterogeneously integrate non-integrable media (atoms, gain media, fluids, biological samples) with PICs. By turning the external medium into a "re-entrant functional element" rather than a terminal stage, the architecture allows the returned signal to re-enter the full on-chip toolbox for filtering, interferometry, and feedback.

The authors emphasize that this approach avoids the need for complex on-chip patterning of external materials or the use of weak evanescent fields. Instead, it utilizes a standardized, millimeter-scale "GC-mirror-GC" footprint that allows external modules to be mounted, replaced, and reworked on a photonic motherboard similar to discrete components on an electronic circuit. The successful demonstration of a chip-scale atomic frequency reference without free-space frequency-conversion optics validates the potential for portable optical atomic clocks, quantum memories, and integrated frequency-comb spectroscopy using this topology.

Technical Summary: A Re-entrant Chip-Free-Space Photonic Interface for Telecom-to-Rubidium Spectroscopy

More like this